Here, take this pill
"Will it cure me?"
You won't have cancer anymore, that's for sure!
"Welp, down the hatch!"
.....
dies
this post was submitted on 09 Jul 2025
110 points (89.3% liked)
Technology
72729 readers
1813 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related news or articles.
- Be excellent to each other!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, this includes using AI responses and summaries. To ask if your bot can be added please contact a mod.
- Check for duplicates before posting, duplicates may be removed
- Accounts 7 days and younger will have their posts automatically removed.
Approved Bots
founded 2 years ago
MODERATORS
can we really trust a “black box” algorithm with our lives?
No. That’s why we have clinical trials.
We don't even trust experts without testing
We don't even trust experts without testing
We certainly didn't but I think the time of don't is long past.
There really needs to be a rhetorical distinction between regular machine learning and something like an llm.
I think people read this (or just the headline) and assume this is just asking grok "what interactions will my new drug flavocane have?" Where these are likely large models built on the mountains of data we have from existing drug trials
Life sciences are where this sort of thing will shine.
Reproduceability is always an issue.
?
Reproducibility of what we call LLM 's as opposed to what we call other forms of machine learning?
Or are you responding to my assertion that these are different enough to warrant different language with a counterexample of one way in which they are similar?
Those models will almost certainly be essentially the same transformer architecture as any of the llms use; simply because they beat most other architectures in almost any field people have tried them. An llm is, after all, just classifier with an unusually large set of classes (all possible tokens) which gets applied repeatedly
A quick search turns up that alpha fold 3, what they are using for this, is a diffusion architecture, not a transformer. It works more the image generators than the GPT text generators. It isn't really the same as "the LLMs".
I will admit didn't check because it was late and the article failed to load. I just remember reading several papers 1-2years ago on things like cancer-cell segmentation where the 'classical' UNet architecture was beaten by either pure transformers, or unets with added attention gates on all horizontal connections.
I skimmed the paper, and it seems pretty cool. I'm not sure I quite follow the "diffusion model-based architecture" it mentioned, but it sounds interesting
Diffusion models iteratively convert noise across a space into forms and that's what they are trained to do. In contrast to, say, a GPT that basically performs a recursive token prediction in sequence. They're just totally different models, both in structure and mode of operation. Diffusion models are actually pretty incredible imo and I think we're just beginning to scratch the surface of their power. A very fundamental part of most modes of cognition is converting the noise of unstructured multimodal signal data into something with form and intention, so being able to do this with a model, even if only in very very narrow domains right now, is a pretty massive leap forward.
I'm not talking about the specifics of the architecture.
To the layman, AI refers to a range of general purpose language models that are trained on "public" data and possibly enriched with domain-specific datasets.
There's a significant material difference between using that kind of probabilistic language completion and a model that directly predicts the results of complex processes (like what's likely being discussed in the article).
It's not specific to the article in question, but it is really important for people to not conflate these approaches.
Actually I agree. I guess I was just still annoyed after reading just previously about how llms are somehow not neural networks, and in fact not machine learning at all...
Btw, you can absolutely finetune llms on classical regression problems if you have the required data (and care more about prediction quality than statistical guarantees.) The resulting regressors are often quite good.
I mean I hate AI in general.. but to be honest... assuming no one is stupid enough to bypass the trials etc... I'm all for it, 90% of these problems already exist in the existing system, who owns it, can a corporation charge us to death.
The only reasonable fear is, if they come out with more than they can develop trials for, and they lobby to lower standards in trials. Even that honestly is a more acceptable risk in the context of terminal diseases/severe cancers.
Agreed, drug development is a very good use of AI.
Rfk Jr will use it for drug approvals too! Uh oh.
and it's still a better system than anyone hired by RFK Jr manually reviewing the file.
Which is kind of the point, idea fully agreed there's a lot of risks and messed up stuff, but almost all of it, is at worse roughly equal to the already existing problems in our systems... I can't quite think of any that are made worse.
assuming no one is stupid enough to bypass the trials etc…
Of course.
There's a separate AI model dedicated to running the trials. Don't worry.
Okay here we go guys! Drink up!
Feel anything yet? Let's try another... Hold up! Wow, I can see 360!
Dude! Gnarly! You got eyes in the back of y.....dude I can see 360 too!
Nah, I only see 360 pills. Where do you see the other two?
Holy wakamoly! You got 360 eye balls!
Experiment 00000000001... Failure...
Did the machine test its first human drug yet?
Yes! LSD!....didn't work.
Well Dr. Chich what will we try next?
We? Or the machine?
I’m sure all the savings from accelerated/cheaper R&D will be passed on to the consumer…right?
They will, just not in the US lol
There’s only one way to solve all diseases.
Did they test this on Mars first?
Sure it helps with a bottle neck but it is not the only one. Until you gain biological and biochemical understanding of the disease no amount of throwing neural networks will help you. I am really sick and tired of AI people hyping up their stuff to get more investments. It even feels like all this "secretive" bullshit is also a part of the show.
Big Pharma says no the hell you aint
Lovely. Let me pencil “zombie apocalypse” back onto my 2026 BINGO card.
I think I lost neurons reading this. Other commenters in this thread had the resilience to explain what the problems with it are.
Is this how we all get AIDS?